GLM-4.1V-9B-Thinking
Open 9B vision-language reasoning model said to match or beat Qwen-2.5-VL-72B on 18 benchmarks, with 64K context.
Adds chain-of-thought plus reinforcement learning to a small VLM, handling images up to 4K resolution at any aspect ratio. First of Zhipu's GLM-V line; succeeded by GLM-4.5V and GLM-4.6V.
- Date
- Tuesday, 1 July 2025
- Lab
- Zhipu AI / Z.ai
- Kind
- open-weights
- Access
- open weights
Figures
| Measure | Value | Measured by |
|---|---|---|
| Benchmarks matching/beating Qwen2.5-VL-72B | 18 tasks company-claimed | company |
GLM-V repo news dates it 2025-07-01; the GLM-4 repo lists 2025-07-02.
Sources
This record was checked against its sources on 6 October 2026. How we check